The minutes for each meeting of the Data Ethics Advisory Group, grouped by date, are summarised here.
Meeting summaries from 2019 to 2022
Advisory Group Members: Professor Colin Simpson (Chair), Jonathan Godfrey, Jonathan Kilgour, Will Koning, Andrew Sporle, Kate O’Connor, Russell Craig
Stats NZ Officials: Emma MacDonald, Director, Stats NZ; Fiona Wharton, Principal Advisor, Stats NZ; Fiona Sinclair, Senior Governance Advisor, Stats NZ.
The Chair welcomed members and officials, and the meeting was opened with karakia.
The minutes of the meeting held 22 August 2024 were approved with minor updates
Members declared relevant interests.
In 2021 the Public Service Commission (PSC) ran the first survey of public servants. The outcome of which provided significant input into the pay gap work led by the PSC.
Key topics of discussion were:
There is a large amount of data coming from agencies ahead of the survey and DEAG queried what was the purpose of this? PSC acknowledged that a large amount of workforce variable data is being received ahead of the survey. This is preferable to requesting the information after the survey for those who participated, as this means that leaders have no visibility of who answered the survey and cannot pressure those that do not respond. In the previous survey where salary, title and role tier information was requested from individuals in the survey, the quality of the data received was low, and could not be used. Quality workforce data is better sourced directly from agencies.
A group member proposed additional information that could be included on the Information Sheet and Consent form to better support informed consent.
PSC indicated that they would make changes to the participant information sheet and informed consent.
In the last six months many in the public service have been made redundant. Is there scope to hear from them as a part of this process? PSC noted that this would fall outside the scope of this survey, however they are aware that the Public Service Association (PSA) is doing some work in that space.
A group member asked how the questions relating to disability and gender were developed and whether participants know that they may be asked these questions before beginning the survey? Assurance of anonymity is needed as survey participants may choose to provide details that they would not like, and have not included, in their agency’s HR record.
PSC noted that a working group was set up to develop the disability questions for this survey. The group included We Enable Us, the Ministry of Social Development, and Whaikaha – Ministry of Disabled People. PSC also noted that the participant information sheet provided with the survey invitation includes details on sensitive topics, and further details on confidentiality.
The overall cost of engagement for this survey will be quite high given the number of public servants and the time taken to respond, will the outcome reflect this?
PSC are encouraging the centralisation of engagement surveys within the public service to ensure transparency, and less cost overall to individual agencies, alongside the ‘engagement cost’ lowering.
A DEAG member spoke to the need to communicate back to participants the outcome of the survey and what is going to be done as a result of the data received. Examples were given of the Project MailCore, and a recent Heather Worth Survey, that gave feedback on distinct issues, building trust within the communities surveyed.
PSC noted that reporting at both the agency and overall public service level were published from the 2021 survey, as well as deep dive research about communities within the public service (women, disabled public servants, rainbow communities). For this 2025 survey, insights will be made publicly available.
A group member noted that the response rates for the survey are not those of a significant magnitude. Understanding those that are not responding to the survey will be interesting in who are they and why are they not responding to the survey. These could represent the ends of the trust spectrum.
PSC noted that as a voluntary staff survey, 63% was a reasonably high response rate of a population. In 2021 responses were lower from those public servants who don’t have desk-based jobs. Analysis comparing the Workforce Data (100% of employees of departments and departmental agencies) with Te Taunaki Public Service Census results will be conducted to identify any patterns of non-response.
Wider conversation was held on current concerns regarding AI and guidance available. Concerns discussed included:
Advisory Group Members: Professor Colin Simpson (Chair), Pieta Brown, Jonathan Godfrey, Jonathan Kilgour, Will Koning, Kate O’Connor, Russell Craig
Apologies: Frith Tweedie, Andrew Sporle
Stats NZ Officials: Emma MacDonald, Director, Stats NZ; Fiona Wharton, Principal Advisor, Stats NZ; Fiona Sinclair, Senior Governance Advisor, Stats NZ.
The Chair welcomed members and officials, and the meeting was opened with karakia.
The minutes of the meeting held 25 July 2024 were approved.
Members declared relevant interests.
Stats NZ Officials noted the IDI has been in place for 15 years, and Stats NZ is committed to protecting the data and making it accessible to approved research projects to maximise value. Government settings will also look at using this data in relation to social investment. Improvements in accessing data, and reducing barriers to access data, are being considered alongside data protection and maintaining trust and confidence. Key Topics discussed were:
The physical and geographic location of the data labs was discussed for access barriers. What are the options for working with the IDI in an open plan office, or other location, and what would this look like in terms of the ‘Five Safes’ framework?
The three main user groups of the IDI are Academia, Government Agencies, NGO’s and Private Consultants who all have different ways of working. Stats NZ can secure the data from an IT perspective, but individual behaviours are harder to control. Currently researchers access the IDI from a data lab, with controls in place in the physical environment that align with Stats NZ’s requirements.
The Administrative Data Research UK SafePod network was noted as a potential short-term, physical access solution in smaller regions. The SafePod (or similar physical safe haven access system) may also be an avenue to closer international working e.g. Safepods allow access to data hosted by several organisations based in multiple UK countries.
Technical and expertise barriers are present in accessing the IDI. Most of the users either have high level qualifications or are in the process of undertaking high level qualifications. The Group Project Pathway was introduced to try to ease these barriers but has not yet seen many applications. The Stats NZ Team outlined other initiatives in place to lower entry barriers.
The role of university ethics bodies in the approval processes for accessing the IDI was discussed. A group member cautioned Stats NZ against relying too heavily on the approval of university ethics bodies, as the standard to which projects are held, can vary by university.
Stats NZ Officials discussed the process that projects must go through for approval to access the IDI. Stats NZ tries to support researchers through this process to arrive at a viable project rather than declining applications.
The DEAG noted that if there was any increase in physical or digital access to the IDI then more resources may be required internally to support the approval process.
The risk appetite within Stats NZ was discussed and any potential changes would prioritise keeping the data safe.*
Methods were discussed for ensuring safety of the data, including the opportunity to provide derived data products rather than accessing data directly, with varying levels of granularity and security based on the level of access permitted. Derived data products could also make the data more accessible.
Approaching larger suppliers - Microsoft and Amazon Web Services - was discussed, as they have significant experience in making data secure and accessible, and could be useful in the development of work on the IDI or the Integrated Statistical Data System.
Members queried if the current model was one of “Zero Trust”, in the overall approach to data security. The team responded that all researchers are vetted before being given access and that security comprises a mix of technical, legal and behavioural protocols. Members encouraged the Stats NZ team to research more modern technologies and data governance tools, such as homomorphic encryption. Caution is also needed to ensure that any additional risks are worth the additional benefits.
Note: This meeting took place before the outcome of the Misuse of Census 2023 data investigations were known. Risk appetite within Stats NZ may be materially affected by these findings and actions.
A group member encouraged undertaking work on engaging with potential users to better understand what needs are not being met and clarifying what the problem is to be solved. Opening up additional access does not improve the ability of potential users to make use of the access, as technical skills are still required. Derived data products may change the areas experiencing difficulties.
A group member noted that identifying need is important. When Te Whata was developed, this made data more accessible and provided the ability to design data for Iwi Māori development. This is as an important function for the IDI. However, there was insufficient data literacy in some communities for meaningful uptake. The education piece for communities is fundamental.
A group member acknowledged that the new Data Lab established in Rotorua was a great initiative but queried the plan to support equity and access into the IDI for other priority groups?
A group member queried if they have considered synthetic data, for use in non-rigorous or academic queries, as this may be sufficient for many users.
Stats NZ Officials noted that they have been considering synthetic data and different level of synthetic data, but maintaining real relationships in the data is complicated. An area where it could be considered useful is in testing code.
The Virtual Health Information Network provides a practical course to introduce prospective researchers to the IDI, using fake data. Demand has been increasing and if access to the IDI is widened, then the course would also need to scale up.
Synthetic data and Digital Twins could provide high fidelity opportunities for ‘high-end users’ to do statistical modelling outside of a data lab. Modelling that is not easily done in the lab and uses high processing power. Users could then schedule the syntax to run out of Data Lab hours.
Stats NZ agrees that there is potential to replicate the physical and technical space of the Data Lab environment using synthetic data. Users could ideate and refine their approach in this environment to lower the amount of time needed within the real Data Lab.
Privacy of the individual and the privacy of the group was discussed in relation to synthetic data creation and use. This is an area that requires careful consideration. One approach would be to use data tagging to restrict some queries.
* Note: This meeting took place before the outcome of the Misuse of Census 2023 data investigations were known. Risk appetite within Stats NZ may be materially affected by these findings and actions.
Advisory Group Members: Professor Colin Simpson (Chair), Pieta Brown, Jonathan Godfrey, Jonathan Kilgour, Andrew Sporle, Russell Craig
Apologies: Kate O’Connor, Frith Tweedie, Will Koning
Stats NZ Officials: Emma MacDonald, Director, Stats NZ; Fiona Wharton, Principal Advisor, Stats NZ; Fiona Sinclair, Senior Governance Advisor, Stats NZ.
The Chair welcomed members and officials, and the meeting was opened with karakia.
The minutes of the meeting held 27 June 2024 were approved.
Members declared relevant interests.
Details of this discussion are currently withheld as the matter is under active consideration. They will be published when final decisions about the next census are confirmed.
Advisory Group Members: Professor Colin Simpson (Chair), Pieta Brown, Jonathan Godfrey, Jonathan Kilgour, Will Koning, Andrew Sporle, Frith Tweedie, Russell Craig
Stats NZ Officials: Emma MacDonald, Director, Stats NZ; Fiona Wharton, Principal Advisor, Stats NZ; Fiona Sinclair, Senior Governance Advisor, Stats NZ.
The Chair welcomed members and officials, and the meeting was opened with karakia.
The minutes of the meeting held 23 April 2024 were approved.
Members declared relevant interests.
Details of this discussion are currently withheld as the matter is under active consideration. They will be published when final decisions about the next census are confirmed.
The Ministry of Business, Innovation and Employment (MBIE) are seeking feedback on what to include in an AI Roadmap* and guidance on risk management*. The Data Ethics Advisory Group members feedback and discussion focussed on:
Understanding the work that is going on internationally within this space will be key. A group member noted that current international interests are focussed on data provenance and auditing capabilities. There is significant work being undertaken out of the Premier’s office in New South Wales, Australia and conversations with them were encouraged, to identify areas where pain points could be avoided.
Understanding was sought on how the risk management guidance will cover the principle of ‘doing good while doing no harm’. MBIE noted that the government recognises AI as an opportunity and wants to embrace innovation to improve lives, while managing risks. There are some mechanisms within the current legislation and more may be needed.
A Group member queried how risk management will provide business certainty when it is only voluntary. The member pointed to challenges faced in a business context in choosing between following best practice and the pursuit of commercial goals.
A Group member identified that the risks surrounding AI may not be shared equally amongst all citizens and populations and cultural differences need to be acknowledged.
A group member noted that only talking to people with good intentions when using these tools would not help progress in understanding deeper risks. Generative AI and the potential to create fake information at scale, and AI application to health data were seen as key concerns. MBIE were encouraged to engage with different sectors of the business community as part of this work.
The establishment of urban myths around the naming of AI products, long-term use of these products, and correction of the urban myths was discussed. Protections for those using and working with data in this space are not yet established. Moral responsibility is needed as technology can be misused.
A member noted that support regarding appropriate data governance should be made clear, as this is poorly understood within NZ and may undermine efforts within AI implementation.
A Group member noted that AI skills training and awareness would be needed for executives and leadership teams, and for workforces that undergo significant changes due to AI and automation. Resources are needed to demystify AI and engage people, with a focus on data ethics.
Cyber-security and resilience within small businesses was discussed, and the potential to lose awareness of risks due to technical capabilities. It was acknowledged that many businesses within New Zealand are sole traders that will likely consume AI in a ‘Software as a Service’ approach.
Members also emphasised the critical importance in AI of data standards, data quality, data governance, data protection and privacy. It was noted that AI is currently at the peak point of the Gartner Hype Cycle and that we will likely be on the other side soon.
* MBIE advise that the names of these documents are subject to Ministerial Agreement and will change.
Advisory Group Members: Professor Colin Simpson (Chair), Pieta Brown, Jonathan Godfrey, Will Koning, Andrew Sporle, Frith Tweedie, Kate O’Connor, Russell Craig
Apologies:
Jonathan Kilgour.
Stats NZ Officials: Emma MacDonald, Director, Stats NZ; Fiona Wharton, Principal Advisor, Stats NZ; Fiona Sinclair, Senior Governance Advisor, Stats NZ.
The Chair welcomed members and officials, and the meeting was opened with karakia.
The minutes of the meeting held 28 March 2024 were approved.
Members declared relevant interests.
Members were notified of the public consultation and invited to submit individual responses if they were interested. It was noted that the interim Centre for Data Ethics and Innovation (iCDEI) will prepare a written submission. ICDEI were invited to circulate this to DEAG members.
Stats NZ opened the discussion on AI procurement in the new AI Impact Assessment Tool.
The Government Chief Digital Officer will write to all government agency chief executives with guidance on Artificial Intelligence (AI) Procurement, and there is an opportunity for iCDEI to include a short note regarding ethical considerations within this.
Initial concerns identified by iCDEI include:
A member noted that the World Economic Forum provided advice on public sector procurement of AI in 2020, as well as the more recent private sector procurement guidance in 2023, and while it won’t cover generative AI it will be a good reference point. Other international examples were noted, including the guidelines within the United Kingdom for AI procurement by government departments and the Responsible AI framework in Canada where AI suppliers need to be approved to supply AI to the government. A group member noted that advice came out overnight from the International Science Council on the use of AI and emerging technologies:
Key areas identified by the group for advice to be developed on were:
Group members emphasised the value of storytelling/sharing case studies to bring potential risks to life e.g., Horizon - Postmaster saga, Robodebt, Netherlands Child Welfare. They noted the importance for AI to be procured only when it will provide the best solution to the problem and noted potential parallels to the advice previously prepared by DIA on the use of Offshore Cloud Services which provides guidance on jurisdictional risks. Additional areas of security concern lie in the rule of law within the countries of overseas suppliers. History of suppliers’ response to jurisdiction issues will inform how New Zealand organisations might be protected.
The group stressed the importance of ensuring that both the procurers and users of AI understand their responsibilities, and that not all responsibility lies with the supplier. Members discussed the risks of automation bias, and need for agencies to train staff involved in monitoring AI decisions. The group also emphasised the importance of explainability, since Government automated decision-making can significantly impact the public, and understanding how decisions are made will allow the public to challenge decisions that unfairly impact them
There is an opportunity for the Government to influence and encourage appropriate AI governance by suppliers through the procurement process. For example, requiring vendors to demonstrate how they are taking steps to manage accuracy, privacy, bias and explainability risks will help encourage a greater focus on minimising the well-established risks of AI.
A three-phase consideration during procurement was proposed:
New versions of an AI model/tool being released post implementation that were trained on new data sets would not change the governance approach as governance standards are irrespective of the AI/software.
Cross-agency consistency in pre-procurement, ethical assessment, and approval processes is essential, especially in the delegation powers for microdata access under the Data and Statistics Act 2022 and potential uses of AI over microdata in different environments.
A group member raised the need to identify potential taxation opportunities on the use of AI by overseas companies. The group discussed the struggle for New Zealand AI companies to compete with overseas companies who benefit from economies of scale and potential areas of focus for New Zealand such as local validation of AI models and tools, and supporting ‘grow your own’ New Zealand AI tools that may have a competitive edge.
Details of this discussion are currently withheld as the matter is under active consideration. They will be published when final decisions about the next census are confirmed.
Advisory Group Members: Professor Colin Simpson (Chair), Pieta Brown, Jonathan Godfrey, Jonathan Kilgour, Will Koning, Andrew Sporle, Frith Tweedie, Kate O’Connor, Russell Craig
Stats NZ Officials: Emma MacDonald, Director, Stats NZ; Fiona Wharton, Principal Advisor, Stats NZ.
The Chair welcomed members and officials, and the meeting was opened with karakia.
The minutes of the meeting held 22 February 2024 were approved.
Members declared relevant interests, it was noted that Will Koning would leave the meeting for Topic Two due to a conflict of interest.
The Action register was discussed.
Stats NZ thanked DEAG for feedback on the draft public consultation questions and advised of a proposed delay to public consultation to allow for additional translations of consultation documents.
An update on engagement with agencies, priority populations and community groups was provided. A committee member was pleased to hear that Stats NZ was considering delaying the public consultation to improve accessibility and assist in getting the breadth of engagement needed.
Feedback from committee members was provided on a consultation question that asked respondents to rank what was most important to them, to include privacy in the list. Feedback also highlighted the need for clear messaging as part of the consultation that Stats NZ’s proposed use of admin data is for the Census only, at a single point in time.
A query regarding consultation with Chinese ethnic groups noted no specific Chinese community engagement was underway; officials noted that a meeting was scheduled with Asian Family Services and attendance at an all-of-government meeting on ethnicity standards would occur.
A committee member noted appreciation for the wording regarding the ‘misrepresentation, under-representation, or not being present in’, administrative data sources - for communities, and recommended that ‘migrants’ be added into this section, as well as this wording being included within the summary document.
A committee member recommended that Stats NZ consider highlighting within the introductory sections, the legislative obligations of Stats NZ to administer the Census, and on citizens to complete. This is something that many of the public may not be aware of.
A committee member noted opportunities to demonstrate on an individual level, how someone might benefit from the use of Census data.
NZTA-Waka Kotahi is taking over the authority to run the safety/speed cameras which are currently operated by Police.
NZTA are investigating establishing an internal digital and data ethics committee and are seeking advice from DEAG on areas of greatest concern to focus the committee on. Questions have also arisen within NZTA regarding Māori Data Sovereignty.
A committee member noted the importance of Māori data being defined by Māori, and the already established Māori data definitions from Te Kāhui Raraunga. It was noted that the definition NZTA are proposing has been adapted from Te Mana Raraunga, and that this is a starting point that will be built on as internal policies develop.
It is likely that safety cameras will collect personal information, particularly where video cameras are used, and so the Privacy Act will apply.
Members discussed their views on what they thought would be key issues for a digital and data ethics committee to focus on. Core issues were identified as being:
A committee member noted that other jurisdictions have enacted specific regulation for ANPR usage, particularly by the Police.
The establishment and structure of the ethics group was discussed. DEAG recommended that the ethics group be external to NZTA, or with a bias to external membership. They also recommended that the data ethics issues related to the operation of the SCS be separated from Māori Data Sovereignty issues to ensure that both topics receive appropriate focus. Members recommended NZTA to remain in contact with the Centre for Data Ethics and Innovation.
Advisory Group: Professor Colin Simpson (Chair), Pieta Brown, Jonathan Godfrey, Jonathan Kilgour, Will Koning, Andrew Sporle, Frith Tweedie, Kate O’Connor, Russell Craig
Stats NZ Officials: Emma MacDonald, Director, Interim Centre for Data Ethics and Innovation, Stats NZ; Fiona Wharton, Principal Advisor, Stats NZ; Amelia Campbell, Principal Policy Advisor, Stats NZ; Fiona Sinclair, Senior Governance Advisor, Stats NZ.
Item One: Stats NZ
Item Two: Te Whatu Ora
The Chair welcomed members and officials, and the meeting was opened with karakia.
The minutes of the meeting held 23 November 2023 were approved. Members declared relevant interests, it was noted that Colin Simpson would step out for the Te Whatu Ora Item and Jonathan Godfrey would Chair for this portion of the meeting. Actions from the previous meeting were reviewed.
The Stats NZ team updated the Group on the response to their feedback. The feedback resulted in Stats NZ beginning to review the wider authorising environment, reviewing legislation, policy, and governance, along with international approaches relating to the use of administrative data. Privacy Impact Assessments are being planned for both the multi register integration and the development of an administrative population model prior to engagement with the Office of the Privacy Commissioner and wider engagement.
Engagement was briefly discussed, noting that those who are most impacted are those that Stats NZ most needs to hear from and where engagement effort should be targeted, rather than engaging with those who are ‘easy to find’. Officials noted that a written update would also be provided to members.
Stats NZ provided an overview to the paper provided and discussion was held with the members.
Engagement is not starting anew, and the team are building on the relationships and partnerships forged through the 2023 Census engagement. Relationships have also been developed through Stats NZ’s Māori Partnerships and Engagement Team, who partner with iwi and Māori to understand their data aspirations and needs.
The Stats NZ team are looking at this through a ‘trust and confidence’ lens and are seeking to find out about how people feel about the use of deidentified person-centred administrative data for official statistics and approved research.
A phased engagement is proposed with key groups. A whakapapa engagement approach with iwi and Māori is planned, including engagement with individual iwi and some collectives that were established as part of the 2023 Census. Phase one is underway and phase two will begin from April. The level of proposed public engagement is new for Stats NZ, where the intent is to engage with the public using a social media/online post-based engagement system, alongside face-to-face engagement with key population organisations and groups, and outreach in community centres / public spaces. Challenges in social media engagement have been recognised and identified.
Timelines were discussed noting that these are very short due to key decision points. A DEAG member noted that this makes the meaningful element of engagement difficult. Discussion was held on the engagement with priority populations, and the identified cohorts.
The Te Whatu Ora team introduced themselves and overviewed their reasoning for approaching the DEAG. Members noted their appreciation for the clear briefing paper provided.
The National Collections team at Te Whatu Ora has systems and processes in place for ensuring that the sharing of the National Collections data they hold is safe and legal. In sharing this with commercial entities there are additional areas regarding ethics and social license obligations to consider that they would like to discuss with the group. Following this meeting they will also being discussing this with the Data Council to hear their views.
Key areas of discussion with the group were:
Te Whatu Ora noted the challenges in resourcing and prioritising requests across the system, and the balancing act required. Additionally, many requests from private entities contain complex technical information. The use of the Health and Disability Ethics Committee (HDEC) as a source of approval for the sharing of information was discussed. The Group were unsure if this was appropriate, and Te Whatu Ora noted that they would approach HDEC with this conversation.
Commercial advantages were discussed, and the group advised that Te Whatu Ora may need to consider comparing providers to identify how they contribute to such things as equity, benefits, and relevance to the New Zealand health system. The importance of fostering local innovation and innovations that don't underserve New Zealanders by using overseas models and algorithms, was also highlighted.
The European Union and Scandinavian approaches to collaboration and cost recovery from researchers and corporate entities was discussed. Noting that cost recovery is a complex field within health, and barriers to individual researchers and smaller research groups could be created.
The Secretariat spoke to the paper provided within the meeting pack, with the proposed changes to the cadence of the Data Ethics Advisory Group.
The members agreed that moving to a monthly, but shorter meeting model was more sustainable and allowed for agencies to approach with items in a timelier manner.
Officials provided an update on engagement with the UK’s National Statistician’s Data Ethics Advisory Committee (NSDEAC) who are keen to engage with DEAG. A pre-recorded presentation has been provided to NSDEAC for their next meeting and in response they will present to a future DEAG meeting.